Tensorflow install on windows

[CameronBodine]

Currently Tensorflow 2.11 does not support GPU on windows. Unsure if this is an issue yet. But when I implement segformer model, I need a Tensorflow version >2.3 as I had errors with predicting using the CPU. It's possible to install Tensorflow=2.10 following the documentation with:

conda create --name ping python=3.9
conda activate ping
conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0
pip install "tensorflow<2.11"

Run the following to see if a tensor is returned:

python -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"

Run the following to see if Tensorflow can find system GPU's:

python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"

This will become relevant with #39 is implemented.

[CameronBodine]

I changed installation of tensorflow-gpu to tensorflow. I haven't been able to predict in parallel using the GPU, so there is no need to have that package. I'm now also installing tensorflow using pip instead of conda. On Windows, this allows the ping environment to using python=3.11 and tensorflow=2.11. This may be modified once I implement SegFormer model architecture for substrate mapping (#39).

[CameronBodine]

Seems to work. Closing the issue.

[dbuscombe-usgs]

famous last words ...